The snowball method is inherently biased due to several factors. Participants volunteering for the sample can create a self-selection error whereby, for example, confident, Englishspeaking individuals may be most keen to participate, risking the exclusion of isolated groups. Those who have received numerous requests in the past may choose to ignore calls for information and a skew in the data may occur because only participants who are currently on the books of gatekeeper organisations may be referred, creating the risk that individuals who have been in the country for longer time periods may be forgotten. Efforts were therefore made to reduce over-reliance on one network, by developing a “multiple access strategy”.20 This helps prevent reliance on particular groups and networks, ensuring instead a range of different potential gatekeepers including statutory organisations, educational establishments, community spaces and religious establishments. Quota sampling It is not possible to draw statistically valid inferences about the whole stateless population through the use of snowball sampling. Rather, this method serves as an indication of potential patterns and relationships. Efforts were therefore made to ensure that participants with certain characteristics were interviewed in order to make the sample as representative as possible of the overall stateless population. Quota sampling was used to implement the non-random selection of respondents according to fixed quotas. The key idea in quota sampling is to produce a sample matching the target population on certain characteristics (for example, by age) by filling quotas for each of these characteristics. It was intended that this method would ensure that the sample reflected key variables and encompassed all relevant groups.21 Quotas were not intended to be too rigid. Sample size and content The agreed aim was for a minimum of 60 per cent of participants to be stateless persons and the remaining 40 per cent or less of participants to be “unreturnable” persons.22 Age and gender were also significant for this study because of the differing international legal obligations owed by the State to children, as well as the potential impact of gender discriminatory nationality laws. The existing data on undocumented migrants (including refused asylum-seekers, overstayers and “unauthorized entrants”) were considered,23 but lacked disaggregation. Consequently, the UK asylum seeking population was put forward as the best basis from which to draw quotas for age and gender. According to statistics published in 2009, women make up around 33 per cent of all asylum applicants. Adults aged 18 to 29 year olds make up over 50 per cent of all applicants, while children under 18 years old comprise just over 10 per cent of all applications. It was hoped to reflect these proportions in the sample. It was also proposed that no single group would make up more than 25 per cent of participants to try to ensure that no profile dominated the sample. There were, however, several key countries of origin and groups that the researchers aimed to cover, namely Kuwaiti Bidouns, Palestinians and British Overseas citizens who had renounced their Malaysian citizenship. It was also hoped to ensure geographic representation 20 18 Snijders, T., Estimation on the basis of snowball samples: how to weight?, in Bulletin de Methodologie Sociologique, 1992. 21 Bloch, A., Zetter, R., and Sigona, N., op. cit. 22 See Chapter 4 for the profile of participants referred and interviewed. 23 Bloch, A., Zetter, R., and Sigona, N., op. cit. Mapping statelessness